車種を判別するAIの作成

#車内判別
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Activation, Dropout, Flatten, Dense
from keras.utils import np_utils
import numpy as np
import keras
import matplotlib.pyplot as plt


classes = ['車内','車']
num_classes = len(classes)
image_size = 64

#メインの関数を定義する
def main():
    X_train,X_test,y_train,y_test = np.load("./car-hanbetu.npy", allow_pickle=True)
    #画像ファイルの正規化
    X_train = X_train.astype('float') / 255
    X_test = X_test.astype('float') / 255
    y_train = np_utils.to_categorical(y_train,num_classes)
    y_test = np_utils.to_categorical(y_test,num_classes)

    model = model_train(X_train,y_train,X_test,y_test)
    model_eval(model,X_test,y_test)

def model_train(X_train,y_train,X_test,y_test):
    model = Sequential()
    model.add(Conv2D(32,(3,3), padding='same',input_shape=X_train.shape[1:]))
    model.add(Activation('relu'))
    model.add(Conv2D(32,(3,3)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2,2)))
    model.add(Dropout(0.25))

    model.add(Conv2D(64,(3,3),padding='same'))
    model.add(Activation('relu'))
    model.add(Conv2D(64,(3,3)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2,2)))
    model.add(Dropout(0.25))

    model.add(Flatten())
    model.add(Dense(512))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(2))
    model.add(Activation('softmax'))

    #最適化の手法
    opt = keras.optimizers.RMSprop(lr=0.00005,decay=1e-6)
    
    #モデルのコンパイル
    model.compile(loss='categorical_crossentropy',
                    optimizer=opt,metrics=['accuracy'])

    #historyに['val_loss', 'val_acc', 'loss', 'acc']を保存
    history = model.fit(X_train, y_train, batch_size=32,epochs=40,validation_data=(X_test,y_test))
        
    #モデルの保存
    model.save("./car-hanbetu.h5")
    
    #学習曲線の可視化
    graph_general(history)

    return model

def model_eval(model,X_test,y_test):
    scores = model.evaluate(X_test,y_test,verbose=1)
    print('Test Loss: ', scores[0])
    print('Test Accuracy: ', scores[1])

    
def graph_general(history):
    # Plot training & validation accuracy values
    plt.plot(history.history['accuracy'])
    plt.plot(history.history['val_accuracy'])
    plt.title('Model accuracy')
    plt.ylabel('Accuracy')
    plt.xlabel('Epoch')
    plt.legend(['Train_acc', 'Test_acc'], loc='upper left')
    plt.show()

    # Plot training & validation loss values
    plt.plot(history.history['loss'])
    plt.plot(history.history['val_loss'])
    plt.title('Model loss')
    plt.ylabel('Loss')
    plt.xlabel('Epoch')
    plt.legend(['Train_loss', 'Test_loss'], loc='upper left')
    plt.show()
        
if __name__ == '__main__':
    main()

#車内判別
from keras.models import load_model
import matplotlib.pyplot as plt
from PIL import Image
import os,glob
import numpy as np
from sklearn import model_selection

img_size = 64
model_param = "./car-hanbetu.h5"

classes = ['4runner','camry','corolla','highlander','prius','rav4','sienna','tacoma','yaris']

model = load_model(model_param)

#それぞれのファイルごとにループさせる
for index, class_ in enumerate(classes):
    photos_dir = "./車/" + class_
    
    #jpg形式の画像データを保存
    files = glob.glob(photos_dir + '/*.jpg')  
    for i, file in enumerate(files): 
        img = Image.open(file)
        img = img.convert('RGB')
        #画像データを64 x 64に変換
        img = img.resize((img_size, img_size))
        # 画像データをnumpy配列に変換
        img = np.asarray(img)
        img = img / 255.0
        pred = model.predict(np.array([img]))
        print(round(pred[0][0] / 1.0, 3) *100 , "%", round(pred[0][1] / 1.0, 3) *100 , "%" )
        ans = np.argmax(pred, axis=1)
        #車外か車内かを判別
        if ans == 0:
          print(">>> 車")
        elif ans == 1:
          print(">>> 車内なので削除")
          os.remove(file)

99.7 % 0.3 %
>>> 車
6.5 % 93.5 %
>>> 車内なので削除

def model_train(X_train,y_train,X_test,y_test):
    model = Sequential()
    model.add(Conv2D(32,(3,3), padding='same',input_shape=X_train.shape[1:]))
    model.add(Activation('relu'))
    model.add(Conv2D(32,(3,3)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2,2)))
    model.add(Dropout(0.25))

    model.add(Conv2D(64,(3,3),padding='same'))
    model.add(Activation('relu'))
    model.add(Conv2D(64,(3,3)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2,2)))
    model.add(Dropout(0.25))

    model.add(Flatten())
    model.add(Dense(512))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(9))
    model.add(Activation('softmax'))

# 確率の平均を取るアンサンブル（ソフトアンサンブル）
def ensembling_soft(models, X):
    preds_sum = None
    for model in models:
        if preds_sum is None:
            preds_sum = model.predict(X)
        else:
            preds_sum += model.predict(X)
    probs = preds_sum / len(models)
    return np.argmax(probs, axis=-1)

# 多数決のアンサンブル（ハードアンサンブル）
def ensembling_hard(models, X):
    pred_labels = np.zeros((X.shape[0], len(models)))
    for i, model in enumerate(models):
        pred_labels[:, i] = np.argmax(model.predict(X), axis=-1)
    return np.ravel(mode(pred_labels, axis=1)[0])